How I Build with AI
Claude Code · Aug 28 – Oct 18, 2026
I run AI agents like a small engineering team. I split the work across parallel sessions, give each a clear brief, check what comes back against real systems, and make the calls that can't be undone myself.
Since August 2026
- sessions
- 185
- commits
- 659
- pull requests merged
- 205
Tools Claude Code · Claude Opus · Claude in Chrome · ChatGPT · iTerm2
Rhythm
2,485 prompts on the 33 days I used Claude in this snapshot.
How I work
-
66 sessions coordinating multiple agents
Coordinate agents in parallel
Backend, mobile and review agents run side by side and pass plans and findings to each other. I decide how the pieces fit together.
-
9,578 shell commands run by agents
Verify against real data
I don't take a fix on trust. Agents check production logs, database queries, simulator runs and the public road data feeds before I treat anything as done.
-
43 decisions Claude handed back to me
I make the irreversible calls
Merges and production access stay with me. Agents bring tradeoffs to me instead of guessing, and ask before touching work they don't own.
-
205 pull requests merged
Ship small, reviewed changes
Work lands as focused pull requests with tests and checks passing. Agents review each other's pull requests, and I regularly strip out comment noise and stale docs.
What the sessions look like
Kinds of sessions
- Multi-step builds 61%
- Iterative refinement 18%
- Focused tasks 13%
- Exploration 5%
- Quick questions 3%
Where agents helped most
- Multi-file changes
- Debugging
- Explaining systems
- Proactive fixes
Currently building
-
TravelDot
A mobile app that turns official road, incident and weather data for your route into a short AI travel summary, so you can decide whether the trip is worth taking before you leave.
-
This site
Static portfolio, including the pipeline that generates this section.
- Update TravelDot to Pre-Launch with site link
- Fix broken light-mode snake image in README (#22)
- Add "How I Build with AI" page generated from Claude Code usage (#21)
Generated from my own Claude Code session data by a small script that runs locally and publishes only aggregate numbers. Updated October 19, 2026.